Predictive Modelling for Skin Disease Classification Using Machine Learning

Skin diseases are among the most common health conditions worldwide. However, diagnosing themaccurately is often challenging due to the complex nature of skin—its texture, color variations, and thepresence of hair can all interfere with proper analysis. To enhance diagnostic accuracy, especially acrossvarious skin types, there is a growing need to develop advanced methods that incorporate machinelearning. Machine learning has become a widely used tool in medical diagnostics, offering promisingresults in identifying skin-related conditions. In this approach, the algorithm processes images of skindiseases and goes through three main stages: feature extraction, training, and testing. By using largedatasets of skin images, the system can effectively train itself to recognize different skin conditions. Thegoal of this process is to improve the accuracy of skin disease classification. Key visual elements such astexture, color, and shape play a critical role in distinguishing between healthy and diseased skin. Inparticular, this study focuses on using color groupings to classify skin conditions. The color of healthyskin differs noticeably from that of affected areas. Similarly, the texture of the skin—whether it's smooth,rough, or patterned—can help in identifying abnormalities. For an effective diagnosis, the studyinvestigates two main types of image features: texture and color. Specific metrics like entropy, variance(dispersion), and HSV (Hue, Saturation, Value) color histogram peaks are used to extract these features.These features are then analyzed using Decision Trees (DT) and Support Vector Machines (SVM), twopopular machine learning algorithms.